840 Pro Speeds

I finally got one shipped in (paid a premium for being in Australia) however... i cloned it from a Dell OEM 830 256GB....
speeds seem a bit off... i have run the Samsung Magician optimizer



any ideas....OS running as normally, used acronis 2013 to clone the drive, any advice

 

Cloning can cause any number of issues.

Why do you think speeds are a bit off?

Different versions of the Intel RST will show different results as may each time you benchmark.

Benchmark first with the MSAHCI driver built into Windows and then with varios RST versions.You may find that the SSD may perform better in benchmarking with the MSAHCI.

Are you benchmarking the SSD while it's the boot drive in use and it has data installed on it?

AS SSD is usually has lower scores than other benchmarks.

Many reviews of SSDs that you may be using for comparison are done on desktop PCs on empty drives benchmarked from another drive that's in Safe Mode.

How does the drive perform in CDM 3.0.2e and Anvil Storage Utilities RC6?

 

Relative to the write speeds, the read speeds seem very low.

Anyhow you replaced a 256GB Samsung 830 SSD, with a 256GB Samsung 840 Pro SSD?

Did you leave any of the 256GB unallocated on the new 840 Pro?

If yes, how much?

Meaning what is/are the partition size/s on the 256GB Samsung 840 Pro SSD?

And how full is the C:\ partition?

 

One note: MS = Microsoft

The whole thing with the MS AHCI driver vs the Intel AHCI driver is like a black hole.

Meaning NBR people mention it, but fail to give insight into what needs to be done.

If you do a clean install of Windows 7 (or Windows 8?), you can avoid loading the MS driver(s), but you have to do it early in the install when you're given the chance to load "SCSI" type drivers, and also have already gotten the appropriate Intel AHCI drivers based on the Intel chipset on your laptop mobo.

1st is that for any normal laptop user, you want the AHCI driver and not the RAID driver.

2nd: download the free info program hwinfo, at: HWiNFO32/64 4.10.1820 Free download - tested and reviewed software downloads from SnapFiles

Run the appropriate 32/64 bit program to find out what your Intel chipset is.

3rd: Then go to the Intel web site and choose "support" and eventually search for: rst

This eventually brings up a long list of Intel's RST drivers.

But now you have the info to pick the right driver.

3a is if you have an AMD cpu, in which case you need to get AMD's AHCI drivers.

 

Man, great post about Intel RST.... it was confusing as all get out when I tried to figure the drivers out! Some say use this version, never that version etc... but I never could find a good post about it... black hole is a GREAT description ha +1

 

No matter whether you have your bios set to AHCI or Raid a Windows 7 or Windows 8 installation will install a basic driver for the storage controller.

You don't need to install anything at the Add Driver screen near the beginning of the Windows installation process.

With the bios set to AHCI the driver will show up as MSAHCI

With the bios set to Raid the driver shows up as Intel v8.6.2.1315

With my bios set to Raid but all 3 installed SSDs set as non member disks Anvil Storage Utilities RC6 shows the Storage Driver as iaStorV.

With my bios set to AHCI Anvil Storage Utilities shows my Storage Driver as MSAHCI.

With Raid configurations it's suggested to match your iRST version to the same like version of the Intel Option Rom of the Raid Utility;that being 11.2 RST and 11.2 OROM.or 11.6 RST and 11.6 OROM.

Your notebooks best SSD benchmarking performance will be found after testing with the drivers embedded in Windows and then various RST versions.

Make sure Write Caching is enabled.

Benchmark after letting the system idle for about 30 minutes.

The amount of data on the drive will effect benchmark performance.

Naturally,your mileage may vary.

 

4k-64 is off numbers are about half. All the tests I've ran came with a result between 1069 and 1074 in AsSSD vs ~620 with the Crucial m4 (no clone clean install)

 

How does Anvil Storage Utilities RC6 benchmark look?
http://www.ssdaddict.com/apps/AnvilBenchmark_RC6.zip

BTW,I saw a new review of the 256GB 840 Pro
http://www.techpowerup.com/reviews/Samsung/840_Pro_SSD_256_GB/1.html

 

I am on my 1st SSD so I am no expert but my 830 is crushing yours on the read 4K-64thread. I get 288MB. I am not far off on the others so my overall read score is 356 yours is 259. You win in writes except a tie in 4K. Access times are close enough to be called the same. The combined final score of mine is 758 vs your 667. What does this mean? I don't really know but the numbers on the read seem off for the new and improved Pro model.

 

the 840 pro has the default allocation...which i presume you mean over-provisioning.......i am afraid i don't understand much about it, but the Samsung software says OP currently is none and recommends i allocate 23.8 GB at-least......but i have 238GB to begin with ...i never did anything anything along the lines of Over-provisioning for the M4 or the 830
Intel RST updated to latest
Running AHCI definetly, i don't do RAID, as i needed to distinct OS. separated from each other so if i select HDD 2 (SSD) on boot-up i go into Win7, and HDD 1 (Seagate HDD) in LEGACY mode (is that an issue as well) i boot up into 8
I paid 248AU$

I have a single partition of 238GB of which 91GB is occupied...full OS running

 

well same OS just increased the power profile to High performance and the result is....


i know about the whole thing on no clone business on the SSD's but guys, at my age i don't have the patience...and with slight OCD characteristics i need everything and every file in the folder as i want it...takes too long nowadays...

would you reckon....just by the way, if i were to use Samsungs data migration software would that help optimise the OS on the 840 during cloning...i used Acronis 2013

 

Looks like working well now and very nice score. I now feel like my 830 is old and slow, oh well.

The correct power profile sure helps.

 

No cloning software can optimize for the O/S (like a clean install of the O/S and programs can). If you're happy with the results you're now getting, leave well enough alone. And yeah the power profile (one reason for a clean install) is not simply setting different timings and percentages - a manufacturer can set any multitude of settings that the user cannot (normally) fiddle with. Glad it helped to use High Performance in your case.


ASUS-UX - don't worry about how the 830 seems old - the 840 Pro is only 2% faster in real world use according to this review:

See:
Samsung 840 Pro SSD 256 GB Review | techPowerUp

 

I've got my cloning down to OCD haha. Can get windows up in like 10-15 minutes and all the programs I need the most are on the same USB flash for easy install/organization. Plus other programs still on the data drive that didn't get formatted.

 

I have no expertise too really comment on this but I must ask with all respect. Cloning at least by definition should not have the issues/drawbacks many keep mentioning, should they? Like I said I lack the expertise to comment point by point but unless cloning is not cloning it should be fine at least in theory. But I do have the expertise to know that many "fresh install" people are guys having fun doing it but also the OCD crowd.

Looks like OP resolved the issue. Scores looked good so I call it a win.

Edit: ewitte I understand your OCD comment and think you really do have the clone under control. I am reffering to that other group that go way beyond the rest of us. Meant no slight or anything else. I myself am OCD about getting my paycheck. OCD is way underated if directed correctly.

 

cant we get scores like these without the high performance profile,
also what about the over provisioning...is that neccessary, and what space should i allocate

 

If you are using the manufacturer's Windows install (and/or the Recovery install...) you simply do not know what 'tweaks' they put into the power options to hit the spec's they promise.

A clean install of a generic Windows O/S is usually the best option to get the most from the platform you bought.

You can see the following link for some of the options that can be tweaked - I have played with them all and found the performance to increase greatly. But so does the heat output and the corresponding decrease in battery life.

See:
http://forum.notebookreview.com/sol...-series-4-5-965-chipsets-stamatisx-tweak.html



As for over-provisioning, I recommend at least 30% additional (not including any built-in spare-area the controller depends on), but Anand suggests at least 25% too:


See:
AnandTech - Exploring the Relationship Between Spare Area and Performance Consistency in Modern SSDs

 

I read your link. Can you explain to me if there is any difference between "spare" area "unused" vs "over provisioning". I can't see so it really does make 90% of what I have heard both ignorant and BS. Not joking, what am I missing?

Don't fill up your SSD beyond 90%? Not complex.

 

If you simply leave some unused capacity on your SSD - some controllers will be able to use it as spare-area (or over-provisioning) - but it is not the same thing as leaving the same unused capacity as 'unallocated'.

The difference is that nand chips used strictly for over-provisioning are 'known' to the controller to be 'clean' and do not need a Read/Update/Erase/Write cycle to be used for the internal GC and TRIM processes the nand requires to stay fast.

If you google my name and davepermen along with notebookreview.com you will see a more detailed answer in a thread here somewhere (davepermen didn't believe it either...).

But unused capacity does no good to the controller for keeping sustained performance over time as high as it can be (it does not 'know' what is real data and what is simply a 'dirty' nand state and therefore it does the R/U/E/W cycle needlessly and to the detriment of the health of the nand - vs. the actual use received).

Unless we're talking using less than 20% of the capacity of the drive, then 'unallocated' capacity used for over-provisioning or extra spare-area duties is highly recommended with current SSD's if the highest sustained performance over time is to be achieved in real world workloads.

 

tiller,

Could you pls provide an actual link. Unfortunately, like myself and davepermen, you and davepermen have many hits where you've disagreed on an issue. I don't have time to go thru every conversation you've had w/ him on "forum.notebookreview.com" as there as just too many.

I'm still skeptical about this. The SSD contains a mapping between LBAs and physical location on the NAND cells. During usage, wear leveling routines will use the ENTIRE inventory of cells available on the drive. All that changes is the mapping of where an LBA can be found in what NAND location. If what you're saying is true, then creating a 10MB partition would cause the SSD to wear out 10MB of NAND cells well before the other ones are even touched (as you're saying the drive won't touch them because it thinks they're clean), and I'm pretty sure that is not the case.

What I believe is the case (let's say for a 256GB drive), is that during use the 10MB is moved around all 256GB in order to wear out NAND cells at an even pace -> and then just change the mapping of where the LBAs can be found in the physical NAND location.

 

You don't have time and I do?

No, wear leveling is different and separate from over-provisioning - we don't know which actual nand chips are put aside, but the controller does and that is the difference with 'unallocated' vs. simply having unused capacity. Where you're going off track is that the same 10MB (I think you mean GB...) of nand is changing from different physical chips over time, depending on the wear leveling algorithm (how aggressive, etc.) and the use the drive is seeing.

If I have time in the next few days, I'll see if I can track down that conversation - but until then, do what I did: read EVERYTHING you can about how SSD's and HDD's work.

(Well, maybe not everything).

 

Okay, forum acting up (or is it Win8? lol...)... I can't edit my post above.

I found the link (well, at least one of the conversations me and davepermen had about this topic):

See:
http://forum.notebookreview.com/solid-state-drives-ssds-flash-storage/619685-ssd-provisioning.html

So, who's skeptical now?

yalcin19 ninja'd me - great information! Thanks. +Rep. Yeah; old news... 2009 - Wow!

 

Happy if I can be any assistance to your case

It should also be mentioned that SSD's treats unused blocks as spare area (ie TRIM).
Having larger spare area helps sustain performance for SSD's in the long run.
Endurance is not an issue anymore as we initially thought SSD's have short lifespan.

 

@tiller,

First off, let me say, no I meant 10MB. If you erase an SSD, and create a 10MB partition the other 255.999 GB (assuming my math is correct) would be unused. Now the drive would allocate a mapping of 10MB of LBAs, but to which NAND cells those go to, it is my guess the SSD will put it anywhere it needs within the "user" available drive space.

I'll go thru the thread you posted. Thanks for finding it.


In regards to the intel post by yalcin19, I still would like to question the author on how "formatting" to something less would work. Use my 10MB example to force an extreme case. Why would the drive not use the ENTIRE drive addressable space when writing for wear leveling purposes. I believe all SSDs already use spare area when writing bytes to the 4K cells, not sure why it would think it could use non-partitioned space for spare area operations.

I guess it *could* be a performance optimization that a controller grabbed some extra NAND cells, and hijacked them for spare area use. But that would really depend on the controller, and if someone expanded a partition, then what?

Also note, "less robust" wording about an under partitioned drive.

 

In the old days, there were SSD's with no TRIM and these SSD's had performance degradation due to write amplification.
And also filling the SSD's near to their capacity cause SSD's performance degradation as well.
Having larger spare area helps SSD's wear levelling even and thus increased endurance of SSD's.

For me the main reason I have allocated a larger spare area is to prevent degradation of my SSD performance and never to worry about it, no matter how rough I use them.

in RAID0 (I am not sure if they have released the drivers for RAID with TRIM yet???) with no TRIM should have large enough spare area to prevent performance degradation in any case.

 

From what I can tell on research, not partitioning is about the same as partitioning and not filling up. The only difference is that computer usage can fill up a fully partitioned drive. In other words, take a newly formatted 10MB partition and add a 1K file, only 1 NAND cell will be allocated (assuming the controller put 1K on the 4K page) the rest of the entire drive is still unallocated and in a clean state. Research suggests that until an LBA is written to, it remains available in the SSD's free pool of clean NAND cells to use as it sees fit.

Addressing some of the comments -> "we don't know which actual nand chips are put aside," -> Answer: None. There is nothing "put aside." The SSD will use ALL space. Just as AnandTech states, "In other words, even if you’re using only 60% of your drive, chances are that 100% of your drive will get written to simply by day to day creation/deletion of files."

Tiller, regarding the other thread with you and davepermen. Thanks for the post. One point to make. It is suggested by to keep 50% of the drive unpartitioned. Now, you know your working patterns best, and I assume that 1/2 spare area works best in that use case. Have you tried anything larger? See page 11 of this IBM report - http://www.research.ibm.com/haifa/conferences/systor2011/present/session5_talk2_systor2011.pdf. The data is a bit stale (Intel X25M G2), but showed data points at 77% over-provisioning. Have you tried anything larger than 50%?

 

I am allocating a further 10% of the space available (after formatting) to O

 

Hey guys,

we are still in the old days of SSD's - today's Intel 525 (mSATA) SSD's are still the same old SF 'garbage' if you don't leave unallocated space to prevent SSD suicide:


Anand Lal Shimpi said:

See:
AnandTech - The Full Intel SSD 525 Review: 30GB, 60GB, 120GB, 180GB & 240GB Tested



Yeah; free space is not equal to unallocated space (yet). No research shows such a thing and my information is from simply using SSD's (on and off) for over 4 years now.

I have tried over-provisioning (via 'unallocated' method) as high as 80% (~95GB out of 512GB Crucial M4) but did not find a worthwhile reason to use the SSD like that (for O/S use OR for PS scratch disk use).


The most bang for the buck is still around 30% over-provisioning (over and above any built-in spare area the particular SSD has).

 

There’s still 7% over-provisioned space from the decimal to binary storage conversion (i.e. GiB to GB). What they’re doing is removing the extra 7% minimum which sandForce had to previously apply on top of this.
So instead of 14% over-provisioning before, it’s now 7% like other manufacturers. That’s why the title says “0%” because it’s not really zero, just the sandForce part is.
In my opinion all SSD NAND blocks should be available to the end user except the blocks reserved for bad block replacement.

 

All do respect tiller you might not know what you are talking about. I really mean it. With free space my SSD performs above specs.

I really mean and have to say. You just don't get it. I don't want to explain the obvious. I get as a rule hardware is better than software. But if conditions allow hardware to work (unused space/trim/reserved) then what advantage is either gained or lost? Teach people with old SSD's and leave the rest of us alone.

 

No, please do tell me the obvious. I may learn something (or you might).

This is not old news what I am reporting here (the SSD's 'tech' we are currently playing with are).

 

While we are on this is LBA0 writes to be concerned about?

 

LBA0? Can you clarify?

 

Yeah it's in the SMART status AIDA64 ..... It shows it has written nearly 360 Gb

 

Sorry. I cannot find any info on what this is about.

 

I have found something regarding Sandforce controllers where they say, just to leave a smaller partition as you cannot fill up the drive in that case, but they do not address a full partition, but not filling it up.

I did come across this quote, which could or could not be true - "There is no notable difference between paritioning a 64GB drive to 50GB and filling up the whole 50GB versus partitioning the drive to its full capacity and only filling it to 50GB. Wear leveling will occur using whatever free space is available, partitioned or not (the process is lower level than the filesystem)." - Crucial C300 and over-provisioning questions - ssd - Storage

Wish I could find something concrete by a manufacturer or controller engineer addressing smaller partition vs. not filling up the partition beyond a certain point. Perhaps this is controller dependent, meaning there is no 'one size fits all' setting, and that is why there's not statement as it all depends on which drive/model one is using?

Well, until then, all I have is real world experience in which my C300 is performing and producing benchmarks just like it did 30 months ago. All this will a full partition, but keeping it under 60% filled.

 

Free space is free space. Unless MFG made with "reserve" it does not matter.

I do hold my notebook up to a full moon twice a year and am certain it helps the SSD performance. I can't prove it.

I do not know the ideal amount to leave free. Seems like 20% is the recomendation. But on my first SSD so no real testing yet.

 

It'd be nice if we could all use the same words.

e.g. there's free space, and allocated/unallocated space, and spare space, and still more space terms.

I went with leaving 20% of my 256GB SSD drive unallocated.

When I get bigger SSD drives (bigger than 256GB), I'll move to 25% unallocated. For smaller SSD drives it is just too painful to not allocate less than 80% of the drive for actual use.

 

20% unused non-allocated space. Wait I put an ambiguous new word in, "unused".

Yea I agree SSD sizing feels a bit painful. Starts lower capacity then we have to not use some of that to maintain performance.

 

Anyone can drive a car without knowing how it works, but it is good to know
when it brakesdown. Even you don't need to fix it on your own, it would help you not getting ripped
off by a mechanic.

 

Let's just talk partitioning then. On my 256GB C300, I went w/ a FULL 100% partition. Have only placed about 130GB of data on the drive. No adverse effects for ~30 months of use.

You have only created partition(s) to fill 80% of the max capacity of your drive. Is that correct?

 

Correct.

The free partitioning program named gparted sees my 256GB Crucial M4 mSATA SSD as 244GiB in size.

I created 1 NTFS partition of 194GiB size, and left 50GiB unallocated.

In another 2 or 3 months I'll get one of the new 480GB Crucial M500 mSATA SSD. AFAIK crucial has already set aside (as unallocated space) at least 32GB.

I expect to keep a grand total of unallocated space on the new drive of about 25%

 

So you would allocate 25% from the total capacity or formatted capacity

 

I don't like picking away at what may seem to be "nits" but you said it in reverse.

I leave 25% of a larger SSD drive unallocated.

 

So 25% of 256gb hdd and then format the remaining space to ntfs